Song J. Kim, MD

Best JG, Barbato C, Ambler G, Du H, Banerjee G, Wilson D, Shakeshaft C, Cohen H, Yousry TA, Al-Shahi Salman R, et al. Association of enlarged perivascular spaces and anticoagulant-related intracranial hemorrhage. Neurology. 2020;95:e2192-e2199.

The clinical significance of cortical superficial siderosis, cortical microbleeds, and cortical subarachnoid hemorrhages — imaging markers of cerebral amyloid angiopathy (CAA) — has become clearer over the years as representing future and recurrent risk of symptomatic intracranial hemorrhages (sICH), especially lobar bleeds. Recently, enlarged perivascular spaces (EPVS) in the basal ganglia (BG-EPVS) and the centrum semiovale (CS-EPVS) have also come to the attention of researchers as possibly linked to hypertensive arteriopathy and CAA, respectively. Given this, could EPVS carry a risk of symptomatic intracranial hemorrhage (sICH) in patients starting anticoagulation? 

To answer this question, Best et al. conducted a post-hoc analysis of the CROMIS-2 (AF) study, a multicenter prospective cohort study examining the relationship between cerebral microbleeds and anticoagulant-related sICH over a 2-year follow-up period. Using baseline MRI imaging pre-anticoagulation, the number of BG-EPVS and CS-EPVS were enumerated and dichotomized (greater than 10 or less than 10). The association of EPVS with symptomatic intracranial hemorrhage was then investigated using Cox Regression analysis, which also incorporated other variables such as cortical microbleeds and lacunes.

Over a 24-month follow-up period, 14 sICH occurred out of 1386 patients — 2 deep, 4 infratentorial, 4 lobar hemorrhages, 2 subdural hematomas, and 1 subarachnoid hemorrhage. In the univariate model, >10 BG-EPVS (HR 10.8 (3.01-38.7), p=0.000), diabetes mellitus (HR 3.88 (1.35-11.2), p=0.012), presence of lacunes (HR 2.97 (1.03-8.57), p=0.044), and cerebral microhemorrhages (HR 3.8 (1.3-10.8), p=0.013) were associated with sICH. In the multivariable model and subsequent sensitivity analysis, >10 BGPVS (8.96 (2.41-33.4), p=0.001) and DM (3.91 (1.34-11.4), p=0.012) remained independently associated to sICH. CS-EPVS did not emerge as an independent variable associated with sICH in either of the analyses.

The authors comment that the association between BG-ePVS, but not CS-ePVS, in anticoagulation-related sICH was an unexpected one, especially as CS-ePVS has been linked to cerebral amyloid angiopathy (CAA). They offer the following explanation: 1) a low proportion of the cohort who met the Boston criteria for CAA (perhaps identified in routine clinical practice and thus not referred to the study) for enrollment; 2) relative sparing of the distal cortical penetrating arteries from higher peak blood pressure compared to lenticulostriate arteries; and 3) low specificity of the association between CS-ePVS and CAA.

Limitations include the low number of adverse outcomes relative to the presence of the markers analyzed, which warrants verification of the findings in future investigations. Despite the significant association found, BG-ePVS does not explain the localization of the sICH outside of the basal ganglia, which occurred in 12 out of 14 cases. Regardless, the strength of the study lies in analyzing pre-anticoagulation imaging biomarkers and the incidence of sICH while on anticoagulation, with findings that EPVS may be more than a benign marker of chronic small vessel disease in the aging population, but a clinically significant one predictive of risk of sICH.